Assessing Physician Response Rate Using a Mixed-Mode Survey

M.D


Introduction
Due to demanding schedules, physicians remain a difficult population from whom to obtain reliable and valid findings. Previous attempts to survey this population using only e-mailed surveys 1 has resulted in low response rates. Using a mixed-mode method for conducting physician surveys may yield higher response rates and reduce the risk of eliminating members of the population who do not use certain technologies regularly. 2 To identify the most effective combination of methods for physician surveys, a literature review was conducted. PubMed, PsychInfo, and Google Scholar were searched for articles with "survey response rate(s)" in the title and were limited to English articles published in the last 10 years. "Physician" was not included to avoid missing articles regarding specialties, such as "family medicine". The search resulted in 149 articles, 103 unduplicated. Titles and abstracts were evaluated by two independent reviewers and articles describing non-physicians were removed (n=70). Full text was reviewed on the remaining 33 articles and 26 were removed for the following reasons: (1) nonphysician respondents, (2) review article, and (3) single method of delivery. Review of the references of the seven remaining articles [3][4][5][6][7][8][9] revealed four additional ones. [10][11][12][13] These 11 articles reported response rates from 32% to 100% (see Table 1). Mixed methods, follow-ups, and incentives may have impacted response rates.
The combination of e-mail, fax, and telephone survey methods has not been reported, based on our systematic review. Therefore, the purpose of this study was to evaluate physician response rates based on this multi-method approach.

Methods
As part of a feasibility study to determine the openness of physicians to using text messages for immunization reminders, a 20-question, IRB-approved survey was conducted with family physicians and pediatricians. Contact information was obtained from the local medical society. Surveys initially were administered by e-mail to 149 physicians utilizing SurveyMonkey © . Two follow-up emails were sent to non-respondents at twoweek intervals. Surveys then were faxed to non-respondents, with a follow-up fax one week later. Finally, phone interviews were attempted with physicians who had not responded by e-mail or fax; each was called twice (Figure 1).
Sixty-seven respondents (66%) indicated current use of a fax machine in their practice, while 94% (96) indicated use of a computer with internet access.

Discussion
Mailed surveys have long been a useful research method with average physician response rates reported in medical journals around 54%. 14 However, mailed surveys can be time consuming and costly, with printing, envelopes, and postage estimated as high as $11 per response. 15 Telephone surveys have shown similar rates and cost per response, 10 but are more time consuming and require an experienced interviewer. 16 Current technology allows researchers to utilize a variety of newer methods, such as fax and e-mail, that are lower in cost and time requirements.
Lensing et al. 17 gave physicians the option of receiving a survey by fax, telephone, or mail. Nearly twice as many physicians requested to be surveyed by fax than other modes, and of those, 87% responded (with fewer follow-ups). Faxing is more cost-effective than mail since no postage is required and surveys can be sent and returned quickly.
The costs of designing and sending a fax survey were estimated as low as $0.52 per response. 18 However, there may be difficulty obtaining fax numbers, loss of anonymity of responders, and possible problems in contact, such as busy signals or inoperable machines. Email surveys can be distributed quickly and inexpensively. Kaplowitz 15 suggested costs as little as $1.32 per response, including computer programming and hosting costs for a web-based survey. Hundreds of surveys can be sent simultaneously by e-mail, while it is often necessary to fax surveys one at a time. Schaefer and Dillman 19 reported returns came in more quickly by e-mail than mail and answers to open-ended questions were more complete. E-mail has similar limitations to faxing, however, physicians also may neglect to access their e-mail on a regular basis, slowing response time.
Assessing the combination of e-mail, fax, and telephone survey methods, after 10 weeks of data collection, our response rate fell within the average range of multimethod surveys of physicians. Our response rate potentially could have improved if more time were allocated to telephone follow-up. However, our intention was to maximize physician responses in a timely and costeffective manner, and follow-up calls were very time-intensive. We, therefore, opted to discontinue after two attempts per person.
Unfortunately, the phone-based studies in Table 1 did not indicate the number of attempts required before physician contact was achieved.
Response rates to phone interviews and fax surveys may have been impacted by clinic staff. Often administrative staff or nurses triage phone calls or fax messages and weed out any that are non-clinical or commercial. Therefore, it is possible some physicians never received the fax or phone message. Other study limitations included: a) the survey was administered in one geographic region; b) no comparison group was used; c) mail and face-to-face interview were not included; d) no incentive was provided, and e) the effects of the questionnaire content, study topic, and question structure were not assessed.
In conclusion, mixed-method survey including e-mail, fax, and telephone followup achieved a reasonable response rate from physicians. Future studies should use randomized groups to assess the cost-and time-effectiveness of different mixed-mode survey techniques. Victorian general practitioners to a mailed survey on miscarriage: Randomised trial of a prize and two forms of introduction to

Introduction
The importance of early identification of weight problems is clear based on the 2003 American Academy of Pediatrics (AAP) Policy Statement on the Prevention of Pediatric Overweight and Obesity, which was also reaffirmed in 2007. 1 The report stated Body Mass Index (BMI) should be calculated and plotted, and urged pediatricians to educate and empower families to raise healthy children. However, studies have shown that physicians do not regularly assess BMI [2][3][4][5] , especially in young children. 5 This is disturbing as physicians tend to underestimate children's weight categories when using only visual cues. 2,6 Early identification and treatment are keys to improving health and decreasing risk of long-term health issues. Many providers are more likely to identify obesity in older children (with higher BMI percentiles), therefore missing opportunities for early interventions. 7 In a study of pediatric subspecialists, children less than five years of age were less likely to receive a diagnosis of overweight. 8 Yet, obesity interventions with children aged 3 to 5 years were more likely to be successful. 9 Therefore, it can be concluded that to treat this medical condition, it is necessary to recognize and identify the problem early.
Parents who were aware that their child's weight was a health problem were more likely to make positive changes than those who did not recognize the problem. 10 Also, parents were more likely to identify weight issues in their child if their doctor had addressed the child's weight. 11 In addition, physician identification of obesity correlated with increased time spent counseling patients. 2 These findings suggested physicians and healthcare workers may have an "ethical obligation" to identifying children who are overweight and sensitively educate parents when treating their overweight child. 10 Correctly identifying and approaching obesity early allows children to be medically managed better. Children clearly identified as overweight in their medical records were more likely to have had laboratory testing than overweight children not identified. 7 In addition, children identified as overweight were more likely to have appropriate medical evaluations for comorbid conditions associated with obesity and earlier interventions for those conditions. 7,8 Because parents have difficulty accurately categorizing children's weight status 6,12 , it is imperative that physicians can identify obese children to counsel parents regarding their child's condition. 10 The problem of pediatric obesity may be common, but it should not be considered "normal".
Physicians should maintain a healthy mental image of appropriate weight and generally be able to identify overweight or obese children by visual cues. This may be especially challenging for resident physicians and medical students, as they represent a group of people who have grown up surrounded by increasing rates of overweight and obesity, therefore, may be desensitized. One study concluded that "children and adolescents who live in environments in which people they see on a daily basis, such as parents and schoolmates, are overweight/obese may develop inaccurate perceptions of what constitutes appropriate weight status". 13 A review of the literature identified no studies that assessed the ability of students or resident physicians to identify weight status accurately. Therefore, this study aimed to evaluate whether community physicians, resident physicians, and medical students are able to categorize preschool children correctly into their appropriate weight class by visual cues alone, and whether there appears to be a difference in accuracy based on training level.

Methods
All fourth-year medical students, family medicine and pediatric resident physicians, and family medicine and pediatric physicians in Sedgwick County, Kansas (a single geographic area), were included in the survey. Institutional Review Board approval was obtained from both the university and hospital.
SurveyMonkey © was used to send an electronic survey to each potential respondent. A reminder e-mail with the link was sent every two weeks to nonrespondents, with a maximum of two reminders after the initial e-mail. Consent was implied by the respondent's willingness to answer the survey.
The survey began by inquiring about the respondent's confidence in his/her ability to identify a child's BMI-for-age-category accurately and beliefs regarding the importance of BMI-for-age status on children's health. The survey proceeded to ask the respondent to rate their perception of the following groups of people on their level of accuracy for visually assessing a child's BMI-for-age: parents, family medicine physicians, pediatricians, nurses, and elementary teachers. Respondents then were asked if they thought different races/ ethnicities perceive children's weights differently, and if so, which race is the least accurate in describing a child's weight.
Pictures of three preschool children were shown. The pictures were the same as ones used in the US Center for Disease Control and Prevention internet-based training on using the BMI-for-Age Growth Charts. 14 The respondent was asked to describe each child's BMI-for-age category using only visual cues (see Figure 1). The weight category choices were underweight (< 5 th percentile), normal weight (5 th to < 85 th percentile), overweight (85 th to < 95 th percentile), and obese (≥ 95 th percentile). The survey concluded by obtaining basic demographic information on the respondents, including age, sex, race, medical training and specialty (or intended specialty if the respondent was a medical student.) Data analysis was completed using the Statistical Package for the Social Sciences (SPSS) version 15.0. Frequencies were computed for categorical data while means and standard deviations were computed for continuous data. Because age was not normally distributed, the median also was reported. For additional analysis, ANOVA was computed to assess total accuracy (sum of correct responses) by level of profession (student, resident physician, or physician) and chi square was used to compare accuracy regarding individual pictures with respondent's gender and confidence level. Due to the low number of respondents for the two extreme categories, the respondents who felt "very confident" were collapsed into "confident" and those who were "somewhat confident" were collapsed into "not confident" for the chi-square analyses to allow Fisher's Exact corrections.

Results
Of the 312 surveys, 134 were completed, resulting in a response rate of 42.9% (see Table 1).
Over 81% (109/134) of respondents indicated they were not confident in their abilities to predict a BMIfor-age category for a child by visual cues accurately, while the remaining 19% (25/134) were confident. Nearly all (95.5%; 127/133) respondents indicated that BMIfor-age status is either important or very important for children's overall health.
Over half of the respondents (51.6%; 67/130) reported that pediatricians visually were able to assess a child's BMI-for-age status accurately, while about one-third of respondents (36.1%; 47/130) indicated that family medicine physicians were able to do so (see Figure 2). Respondents were varied on their confidence in a nurse's ability to assess a child's BMI-for-age status accurately, while they were even less confident in elementary school teachers' abilities. Only 3% (4/134) of respondents reported that parents visually were able to assess their child's weight accurately. Further, 97% (130/134) of respondents indicated that different ethnicities perceive children's weight differently, with Hispanics as the least accurate in describing a child's weight, followed by African Americans.  (3) 7.4% (4) 9.6% (5) Figure 2. Respondents' beliefs in each group's ability to assess a child's BMI-for-age accurately by visual cues alone (as percents).

Discussion
This survey found medical professionals and trainees believe BMI-for-age status is important for health. However, our data showed medical professionals and trainees were unable visually to assess a child's BMI-for-age weight status accurately. At best, classification of children by visual assessment can be described as inconsistent. This supported Barlow's 2007 findings 3 that pediatricians were able to identify only 27% of children with a BMI between the 85 th and 94.9 th percentile. However, the Barlow study also found that 86% of pediatricians identified children as obese if their BMI was above the 95 th percentile for their age, a better rate of recognition than found in our study. An Australian study 2 found general practitioners correctly categorized children into BMI-for-age classes 72% of the time, while pediatricians had a 68% overall correct categorization. Our study contrasted this finding, as there were no statistically significant differences between pediatric and family medicine physicians' abilities to categorize children accurately.
Our results suggested that physicians and trainees recognize the difficulty in assessing BMI category visually, as less than 2% of respondents felt "very confident" in their ability to assess a child's BMI-forage category just by looking at a child. In addition, over half of respondents felt pediatricians would be accurate and less than 40% felt family medicine physicians would be accurate. Our study indicated very low confidence in visual assessment by self or other physicians, which is a clear indicator that the AAP's recommendations for charting BMI are not only appropriate, but necessary. Physicians report barriers to charting BMI, such as time and complexity of explaining BMI 5 , but also endorse innovations, such as color-coded charts, which increase the likelihood of identification and discussion of BMI. 4 In addition, the use of electronic medical records with automatic BMI computations improved documentation of obesity (BMI ≥ 30) by 40%. 15 This study had several limitations. The small area (Sedgwick County, KS) of this cross-sectional survey limited the generalizability of the results. Further, physicians are notably poor survey responders 16 and studies 17,18 report a dramatic decline in response rates of email surveys. However, email was chosen because it is cost-effective, in both time and monetary savings and results in fewer unanswered items than other modes. 19 Our response rate was over 40% after three personalized attempts. According to an article by Schoenman and colleagues 20 , survey results do not differ significantly between 30% and 60% response rates, so we closed survey collection two weeks after the third reminder.
Another limitation is the formatting and form of the pictures. The pictures were black-and-white and respondents may have chosen differently had the pictures been in color. All of the pictures were of non-Hispanic white children, which also could affect perception. Additionally, it is difficult to assess a patient's weight from a photograph, and the children were sitting down which further complicated the assessment. If the children were patients in their own clinics, subjects may have categorized them differently. However, these pictures were used by the US Centers for Disease Control and Prevention for their BMI training and have been used in previously published studies. 21 Further, the physician confidence question was asked before the photos were viewed and responses indicated a dearth of conviction that they would be able to identify any child by sight alone successfully. Finally, the respondents were not asked whether they previously had viewed the pictures on the survey.

Introduction
Amiodarone is a commonly used drug for the treatment of cardiac arrhythmias. It has a broad range of toxicity including photosensitivity, blue-gray discoloration of the skin, thyroid dysfunction, corneal deposits, abnormal liver function tests, and bone marrow suppression. 1 Pulmonary toxicity is the most serious adverse effect of amiodarone. Treatment of amiodaroneinduced pulmonary toxicity includes discontinuation of the drug and initiation of glucocorticoid therapy in the majority of symptomatic patients. This case of recurrent amiodarone toxicity was manifested by acute interstitial pneumonitis and diffuse alveolar hemorrhage after a rapid taper of glucocorticoids.

Case Report
A 60-year-old male patient with no previous lung disease was admitted to the hospital with dyspnea and hypoxemia of 85% on room air. On physical examination, he had bilateral late inspiratory rales involving two-thirds of the chest. His chest x-ray showed diffuse bilateral pulmonary infiltrates (see Figure 1). The patient had no history of occupational pulmonary exposure or history of rheumatologic disorder or joint pain. He was diagnosed with atrial fibrillation nine months prior to presentation and was started on amiodarone 200 mg daily at that time.
After admission, his laboratory workup revealed normal blood counts and electrolytes. His brain natriuretic peptide (BNP) level was 150 pg per milliliter. A 2-D echocardiogram showed normal ejection fraction. There was no clinical or radiologic improvement after empiric antibiotics and treatment with diuretics.
A high resolution computed tomography of the chest was performed and showed bilateral ground glass opacities associated with small bilateral effusions (see Figure 2). Amiodarone-induced hypersensitivity pneumonitis was suspected. Subsequently, amiodarone was stopped and treatment with glucocorticoids initiated. There was a significant clinical and radiologic improvement (see Figure 3) on a follow-up visit one month later. The glucocorticoids were tapered over two weeks, then stopped.
One week after stopping glucosteroids, the patient complained of shortness of breath on minimal activity, rapidly worsening to occur also at rest. He also complained of a dry cough that progressed over the next few days to mild hemoptysis. He again was hypoxic with bilateral rales on physical examination and his chest x-ray revealed bilateral infiltrates. Bronchoscopy was performed and transbronchial biopsies were obtained from the right lower lobe. The pathology examination showed interstitial fibrosis with no evidence of vasculitis or neoplasm (see Figures 4 and 5).
Bronchoalveolar lavage in the right middle lobe yielded a progressively bloodier return. Cytology of the fluid was negative for malignant cells and revealed numerous lipid-laden macrophages (index of 70%), characteristic of amiodarone toxicity. A significant number of hemosiderin-laden macrophages (index of 41%) were present, confirming the diagnosis of alveolar hemorrhage.
The patient was started on intravenous glucocorticoids. He significantly improved and was discharged home on prednisone 60 mg daily with plans to follow-up for several months. A follow-up examination after two weeks revealed clinical and radiologic amelioration.

Discussion
Pulmonary toxicity secondary to amiodarone use occurs in 5-15% of patients. 2 Pulmonary manifestations range from mild to severe and even fatal disease such as ARDS. The most common presentation is interstitial pneumonitis accounting for one-third of patients. 3 Interstitial pneumonitis usually is recognized after two or more months of therapy, especially in patients in whom the dose of amiodarone exceeds 400 mg per day. Cytology from bronchoalveolar lavage is characterized by mononuclear cells predominance and foamy alveolar macrophages. Type II cell hyperplasia and pulmonary fibrosis have been reported. Other manifestations include organizing pneumonia, 4 eosinophilic pneumonia, and lung nodules. 5 Alveolar hemorrhage is a rare complication of amiodarone pulmonary toxicity. Only a few cases were reported in a large study of 171 patients. 4 Two other cases of alveolar hemorrhage secondary to amiodarone use also were reported. 6 Amiodarone is a highly lipophilic drug that avidly binds to adipose tissues, resulting in a large distribution volume and a prolonged half-life reaching 180 days. 7 Given the high accumulation in adipose tissue, pulmonary toxicity may progress despite drug discontinuation. Treatment includes stopping the offending drug with initiation of glucocorticoid therapy in severe cases. Prednisone at 40 to 60 mg daily is recommended with slow tapering, as tolerated, over two to six months. Despite the slow tapering, cases of recurrent pulmonary intoxication may occur. 8 In our case, the rapid tapering of prednisone may have been responsible for the acute recurrence of a more severe form of interstitial pneumonitis with evidence of diffuse alveolar hemorrhage. The patient's elevated body mass index suggested a high volume of distribution of amiodarone which might explain the severity of the recurrent disease. This case emphasizes the importance of slow tapering of glucocorticoids following amiodaroneinduced lung injury.

Use of the QuantiFERON ® -TB Gold Assay in Pregnant Patients
I find the article, "Use of the QuantiFERON ® -TB Gold Assay in Pregnant Patients", 1 in your journal very interesting. I have not read any other work done on QuantiFERON in pregnant woman. In our practice, we have noticed a few pregnant women, especially in second and third trimester, have had "indeterminate" QuantiFERON ® -TB Gold Assay results and when these were repeated at six-weeks postpartum, they showed a definite positive or a negative result. Our feeling is the immune changes during pregnancy are significant in later half of the pregnancy which interferes with the QuantiFERON ® -TB Gold Assay. We would be interested to know what was the gestation of the women enrolled by Chehab BM et al. for their study "Use of the QuantiFERON ® -TB Gold Assay in Pregnant Patients"?